全球高分辨率水储量异常数据集（2002-2022）

陆地水储量异常（TWSA）和地下水储量异常（GWSA）在水文研究和水资源管理中具有重要意义。然而，来自重力恢复和气候实验（GRACE）及其后续任务（GRACE-FO）的产品受卫星设计和不同机构处理策略不同的限制，导致存在多个次优产品。而且，这些产品不仅具有粗糙的空间分辨率，在2017年7月至2018年5月之间还存在连续11个月的间断，给相关研究带来了显著的局限。因此本研究首先基于BTCH（Bayesian-based Three-Cornered Hat）方法融合了多个机构发布的GRACE产品获得了不确定性最低的产品（0.5°），并在此基础上构建了基于物理约束滑动窗口的机器学习降尺度框架（Machine Learning Downscaling Framework based on Physically Constrained Sliding Window，MLDF-PCSW）实现了对应产品空间分辨率的提升（从0.5°到0.05°）及连续缺失月份的填补（2017年7月到2018年5月），并最终生产了一套全球连续数据集，命名为HWSA v1.0（High-resolution Water Storage Anomaly），该数据集包含信息（变量：TWSA和GWSA；空间分辨率：0.05°；时间分辨率：月；时间跨度：2002年4月至2022年12月）。基于三角帽（Three-Cornered Hat，TCH）方法的不确定性分析结果显示，在三个原始产品中，来自美国德克萨斯大学研究中心（Center for Space Research，CSR）的TWSA具有最低的不确定性（25.38 mm），而在基于BTCH方法融合后，不确定性减少了35.07%（16.48 mm）。基于MLDF-PCSW的HWSA v1.0数据集不仅保留了融合的TWSA/GWSA产品的原始信息，而且刻画了更详细的空间细节，与降尺度之前的数据相比在相关系数（Correlation Coefficients，CC）和均方根误差（Root Mean Square Error，RMSE）指标上表现出色（TWSA：CC=0.99，RMSE=0.68 mm；GWSA：CC=0.99，RMSE=1.24 mm），与原位地下水位观测的对比也显示出较高的一致性（最大CC值为0.96）。与其他的重建的TWSA产品的比较结果也表现较好，全球平均CC值达到0.98。显然，该数据集具有为相关研究提供强大数据基础的巨大潜力。

Terrestrial Water Storage Anomaly (TWSA) and Groundwater Storage Anomaly (GWSA) are of great significance in hydrological research and water resources management. However, products from the Gravity Recovery and Climate Experiment (GRACE) and its follow-up mission (GRACE-FO) are limited by differences in satellite design and processing strategies across institutions, resulting in multiple suboptimal products. Moreover, these products not only have coarse spatial resolutions, but also feature an 11-month continuous data gap between July 2017 and May 2018, which imposes significant limitations on relevant research.

Therefore, this study first fused multiple GRACE products released by different institutions using the BTCH (Bayesian-based Three-Cornered Hat) method to obtain the product with the lowest uncertainty (0.5° spatial resolution). On this basis, a Machine Learning Downscaling Framework based on Physically Constrained Sliding Window (MLDF-PCSW) was constructed to improve the spatial resolution of the target product (from 0.5° to 0.05°) and fill the continuously missing months (July 2017 to May 2018). Finally, a global continuous dataset named HWSA v1.0 (High-resolution Water Storage Anomaly) was produced. This dataset includes the following details: variables (TWSA and GWSA), spatial resolution (0.05°), temporal resolution (monthly), and time span (April 2002 to December 2022).

Uncertainty analysis based on the Three-Cornered Hat (TCH) method indicates that among the three original products, the TWSA from the Center for Space Research (CSR), University of Texas at Austin, has the lowest uncertainty (25.38 mm). After fusion via the BTCH method, the uncertainty is reduced by 35.07% (16.48 mm).

The HWSA v1.0 dataset based on MLDF-PCSW not only retains the original information of the fused TWSA/GWSA products, but also captures more detailed spatial features. Compared with the pre-downscaling data, it performs excellently in terms of Correlation Coefficients (CC) and Root Mean Square Error (RMSE) metrics (TWSA: CC=0.99, RMSE=0.68 mm; GWSA: CC=0.99, RMSE=1.24 mm). Comparisons with in-situ groundwater level observations also demonstrate high consistency, with a maximum CC value of 0.96. Comparisons with other reconstructed TWSA products also yield favorable results, with a global average CC value reaching 0.98.

Obviously, this dataset has great potential to provide a robust data foundation for relevant research.

搜集汇总

背景与挑战

背景概述

该数据集提供了2002年4月至2022年12月期间全球高分辨率的水储量异常数据，包括陆地水储量异常（TWSA）和地下水储量异常（GWSA），空间分辨率为1km至10km，时间分辨率为月。数据集通过融合多个GRACE产品并使用机器学习降尺度框架，提高了数据的空间分辨率和连续性，为水文研究和水资源管理提供了重要支持。